Boko Haram Terrorists Exploited AI Chatbots for Bomb-Making and Attack Planning

A recent investigation by The New York Times has brought to light how members of the Boko Haram terrorist group have been leveraging artificial intelligence (AI) tools to enhance their combat capabilities, including the construction of explosive devices and overall operational strategies. The report indicates that former insurgents utilized various AI platforms such as ChatGPT, Claude, Gemini, Grok, and DeepSeek.

Antonia Juelich, a terrorism and technology researcher from the University of Cambridge, was informed by former Boko Haram commanders that following an unsuccessful assault on a military facility, which was thwarted by a defensive trench, the group sought guidance from AI applications. It was through these tools that they discovered methods to modify motorcycles, enabling them to clear such obstacles. Their mechanics subsequently adapted the motorcycles to boost their speed and acceleration, with fighters undergoing practice drills before deployment.

One of the former Boko Haram commanders recounted to Juelich, “We saw in a movie how motorcycles can jump over bridges. We used AI to learn how to do this. We gave it information like which motorcycles we use, the distance we need to jump, and so on. And it gave us steps on what we have to do.”

Another former commander, previously affiliated with the Islamic State West Africa Province (ISWAP), informed Juelich about the ease of use: “You type in the question or use your voice, and it gives you a detailed answer, like ‘How can I build a bomb?’ and then it tells you how. It is like a human robot! We used it a lot. Trial-and-error can kill you. AI gives you accuracy,” he stated. He further added, “Before, the bomb explosion was not that big, but then they studied it. AI told us what chemicals to put in that made the explosion heavier.”

In response to these alarming discoveries, OpenAI communicated to The New York Times that such activities by terrorists constitute a violation of its operational policies. Similarly, Google and Anthropic affirmed that their respective AI models incorporate safeguards designed to reject hazardous inquiries. However, the former insurgents disclosed to the researcher that they frequently circumvented chatbot restrictions by presenting their dangerous requests as legitimate or harmless projects.